Skip to content

Commit

Permalink
Implemented counter-example support in Sly.
Browse files Browse the repository at this point in the history
When conflicts are encountered, Sly will output examples of sequences of symbols and how the parser could interpret them. A couple of examples are provided and a small explanation is added in the documentation.
  • Loading branch information
bugengine authored and yngwe@fry committed Jan 22, 2023
1 parent 4000988 commit ab00964
Show file tree
Hide file tree
Showing 7 changed files with 2,655 additions and 43 deletions.
10 changes: 9 additions & 1 deletion CHANGES
Original file line number Diff line number Diff line change
@@ -1,5 +1,13 @@
Version 0.6
-----------

01/22/2023 Experimental support for counterexamples. SLY will now give
counterexamples for shift/reduce and reduce/reduce conflicts.


Version 0.5
-----------

10/25/2022 ***IMPORTANT NOTE*** This is the last release to be made
on PyPi. If you want the latest version go to
https://github.com/dabeaz/sly.
Expand All @@ -19,7 +27,7 @@ Version 0.5
index of the matching text. This is used to do more
precise location tracking for the purpose of issuing
more useful error messages.

05/09/2020 Experimental support for EBNF choices. For example:

@('term { PLUS|MINUS term }')
Expand Down
108 changes: 108 additions & 0 deletions docs/sly.rst
Original file line number Diff line number Diff line change
Expand Up @@ -1174,6 +1174,114 @@ also be stressed that not all shift-reduce conflicts are bad.
However, the only way to be sure that they are resolved correctly is
to look at the debugging file.

Conflict counterexamples
^^^^^^^^^^^^^^^^^^^^^^^^

To help tracking conflicts, SLY generates counterexamples in the debug file.
For each conflict, SLY will generate one or more examples for each
possibility in a shift/reduce or reduce/reduce conflict. The examples are
a sequence of terminals and nonterminals that the grammar could interpret
in two ways. SLY will show the different derivations, showing clearly what
the ambiguities were. The counterexamples are listed at the end of the debug
file and look like this::

shift/reduce conflict for ELSE in state 11 resolved as shift
shift using rule if_statement -> IF LPAREN expr RPAREN statement . ELSE statement
╭╴
│ IF LPAREN expr RPAREN IF LPAREN expr RPAREN statement ♦ ELSE statement
│ ╰if_statement──────────────────────────────────╯
│ ╰statement─────────────────────────────────────╯
│ ╰if_statement────────────────────────────────────────────────────────╯
│ ╰statement───────────────────────────────────────────────────────────╯
╰╴

reduce using rule if_statement -> IF LPAREN expr RPAREN statement .
╭╴
│ IF LPAREN expr RPAREN IF LPAREN expr RPAREN statement ♦ ELSE statement
│ ╰if_statement───────────────────╯
│ ╰statement──────────────────────╯
│ ╰if_statement────────────────────────────────────────────────────────╯
│ ╰statement───────────────────────────────────────────────────────────╯
╰╴

For each counterexample, the display starts with the list of symbols that cause
an ambiguity. The diamond shows the current location of the parser, and the
symbol following the diamond is the lookahead.
The lines below the symol sequence show the possible reductions according to the
grammar rules. The problem displayed here is the `dangling else
<https://en.wikipedia.org/wiki/Dangling_else>`_ issue;
the first counterexample shows the reduction sequence if the shift path is taken;
the ``ELSE`` is attached to the rightmost ``if_statement``. In the second example,
SLY shows that another interpretation could be to reduce the second
``if_statement`` early and attach the ``ELSE`` sequence to the leftmost
``if_statement`` instead.

Here is an example of a reduce/reduce conflict that occurs in the C language::

reduce/reduce conflict for ) in state 21 resolved using rule expr -> IDENTIFIER
rejected rule (declarator -> IDENTIFIER) in state 21
reduce using expr -> IDENTIFIER with lookahead )
╭╴
│ TYPENAME ( IDENTIFIER ♦ )
│ ╰expr──────╯
│ ╰expr───────────────────╯
╰╴

reduce using declarator -> IDENTIFIER with lookahead )
╭╴
│ TYPENAME ( IDENTIFIER ♦ ) ;
│ ╰declarator╯
│ ╰declarator────╯
│ ╰decl─────────────────────╯
╰╴

In the same way as the shift/reduce conflict, SLY shows here the two ways of
understanding the sequence. It will always backtrack far enough to find the lookahead
symbol after a reduction.

Sometimes, it can be hard to understand why SLY encounters a conflict in the
first place. Consider this example taken from the C11 grammar::

shift/reduce conflict for [ in state 561 resolved as shift
shift using rule attribute_specifier -> . [ [ attribute_list ] ]
╭╴
│ identifier ♦ [ [ attribute_list ] ]
│ ╰attribute_specifier───╯
│ ╰attribute_specifier_sequence╯
│ ╰direct_declarator──────────────────────╯
╰╴

reduce using rule direct_declarator -> identifier .
╭╴
│ identifier ♦ [ ]
│ ╰direct_declarator╯
│ ╰array_declarator─────╯
╰╴

Here, the two sequences are not ambiguous when considered in their entirety;
it is clear that the symbol following the first ``[`` will determine if the
input was supposed to be interpreted as an ``array_declarator`` or as a
``direct_declarator`` including an ``attribute_specifier``. While the grammar
is not ambiguous in this example, it is not LR(1) (in this case, LR(2) would
have removed the conflict) and the state machine does not have enough information
at the time it encounters the lookahead symbol to unambiguously determine what
to do next.

Viewing the state machine
^^^^^^^^^^^^^^^^^^^^^^^^^

SLY can save the state machine in a file in the DOT format that can be used with
graphing tools such as `graphviz <https://graphviz.org/>`_ or even viewed online
in `Edotor <https://edotor.net/>`_. The graph can help visualize how the state
machine is built. Such graphs can become very big for big grammars and are not
always practical, but building such a graph for a small grammar.

In order to generate such a file, add a ``dotfile`` attribute to your
class like this::

class CalcParser(Parser):
dotfile = 'parser.gv'

Syntax Error Handling
^^^^^^^^^^^^^^^^^^^^^

Expand Down
Loading

0 comments on commit ab00964

Please sign in to comment.